1945-1954
Toward a Science of Security Analysis
Benjamin Graham
THE SClEN'I'IRC METHOD
As H. D. Wolfe pointed out in his paper in the last
JOURNAL(Science as a Trustworthy Tool) 1, scientific
method includes among its factors the wide obser-
vation and recording of events, the construction of
rational and plausible theories or formulas, and
their validation through the medium of reasonably
dependable predictions. There are many varieties
of scientific or quasiscientific disciplines, and the
character of the predictions based on them will
vary greatly from one to another.1
At one extreme take the microphone. An
electrical engineer, having rigged it up carefully,
can predict that a word spoken into it will be
immediately amplified. The prediction is precise;
the verification prompt and unquestionable. At the
other extreme let us take psychoanalysis---a disci-
pline sometimes compared with our own security
analysis. Here prediction and verification are less
definite. A layman who finances psychoanalytical
treatment for one of his family is apt to be slightly
in the dark about such details as the nature of the
illness, the method and duration of the treatment,
and the extent of the cure, if any. About the only
thing he can predict with certainty is how much it
will cost per hour. Between these two extremes lies
actuarial science, which to my mind is more rele-
vant than the others to the scientific possibilities Of
security analysis. The life insurance actuary makes
predictions concerning mortality rates, the rate of
earnings on invested reserves, and factors of ex-
pense and profit--in all instances based largely on
carefully analyzed past experience, with allowance
for trends and n e w factors. Out of these predic-
tions, with the aid of mathematical techniques, he
fashions suitable premium schedules for various
types of insurance. What is most important for us
about his work and his conclusions is that he deals
not with individual cases but with the probable
aggregate result of a large number of similar cases.
Diversification is of the essence in actuarial sdence.
Thus our first practical question about "scien-
tific security analysis" is whether it is actuarial in
character, and has diversification as its essential
Reprintedfrom The Analysts Journal (August 1952):97-99.
FinancialAnalysts Journal / January-February 1995
© 1995, AIMR®
ingredient. One plausible answer may be that
diversification is essential for certain types and
objectives of security analYsis but not for others.
Let us classify the things that security analysis tries
to do and see how the element of diversification
applies to each. At the same time we may raise
other questions concerning the scientific methods
and predictions operating in each of the classes.
I suggest that the end product of our work
falls into four different categories, as follows:
1. The selection of safe securities, of the bond
type.
2. The selection of undervalued securities.
3. The selection of growth securities, that is,
common stocks that are expected to increase their
earning power at considerably better than the
average rate.
4. The selection of "near-term opportunities,"
that is, common stocks that have better-than-aver-
age prospects of price advance, within, say, the
next 12 months.
This list does not include stock market analy-
sis and predictions based thereon. Let me com-
ment briefly on this point. If security analysis is to
be scientific, it will have to be so in its own right
and not by depending on market techniques. It is
easy to dismiss this point completely by saying
that, if market analysis is good, it doesn't need
security analysis; and, if it isn't good, security
analysis doesn't want it. But this may be too
cavalier an attitude toward an area of activity that
engages the interest of a host of reputable security
analysts. That stock market analysis and security
analysis combined may be able to do a better job
than security analysis by itself is at least a conceiv-
able proposition and perhaps a plausible one. But
t h e burden is on those who would establish this
thesis to demonstrate it to the rest of us in un-
equivocal and convincing fashion. Certainly the
published record is far too meager, as yet, to
warrant conceding a scientific standing to a com-
bination of the two analyses.
FOUR CATEGORIES
To return to our four categories of security analy-
sis, choosing safe bonds and preferred stocks is
certainly the most respectable if not the most
25
1945-1954
exciting occupation of our guild. Not only has it
major importance of its own, but also it can offer
useful analogies and insights for other branches of
our work. The emphasis of bond analysis is on
past performance, tempered by a conservative
view of future changes and dangers. Its chief
reliance is on a margin of safety that grows out of
a small ratio of debt to total real value of the
enterprise. It requires broad diversification to as-
sure a representative or average over-all result.
These viewpoints have made bond investment, as
practiced by our financial institutions, a soundly
scientific procedure. In fact, bond investment now
appears to be almost a branch of actuarial science.
There are interesting similarities (as well as differ-
ences) between insuring a man's life for $1,000
against a premium of $34 per year, and lending
$1,000 on a long-term bond also paying $35 per
year. The calculated mortality rate for men aged 35
is about 4 out of 1,000, or 4/10% per year. A
comparable "mortality rate" might be applied to
corporate enterprises in the best financial and
operating health, to estimate the risk attaching to
high-grade bond investment. Such a figure, say
1/2%, might then properly measure the risk and
yield differential between the strongest corporate
bonds and U. S. Government obligations.
BOND INVESTMENT: A SCIENTIFIC
PROCEDURE
Bond investment should take on more of the
character of a scientific procedure when the mon-
umental corporate bond study, carried on by the
National Bureau of Economic Research and other
agencies, is finally completed and the mass of
statistical data and findings is made available to
security analysts. The greatest weakness of our
profession, I have long believed, is our failure to
provide really comprehensive records of the re-
sults of investments initiated or carried on by us
under various principles and techniques. We have
asked for unlimited statistics from others covering
the results of their operations, but we have been
more than backward in compiling fair and ade-
quate statistics relating to the results of our own
work. I shall have a suggestion to make on that
point a little later.
SELECTION OF UNDERVALUED SECURmES
The selection of undervalued securities appears
next on my list because of its logical relationship to
investment in safe bonds or preferred stocks. The
margin-of-safety concept is the dominant one in
b o t h g r o u p s . A common stock is undervalued,
,26
typically, if the analyst can soundly establish that
the enterprise as a whole is worth well above the
market price of all its securities. There is a close
analogy here with bond selection, which also re-
quires an enterprise value well in excess of the
debt. But the rewards for establishing that a com-
mon stock is undervalued are, of course, incompa-
rably greater; for in the average case all or a good
part of the margin of safety should eventually be
realized as a profit to the buyer of a truly under-
valued issue.
In this connection I want to throw out a broad
and challenging idea--that from a scientific stand-
point common stocks as a whole may be regarded as
an essentially undervalued security form. This point
grows out of the basic difference between individ-
ual risk and overall or group risk. People insist on
a substantially higher dividend return and a still
larger excess in earnings yield for common stocks
than for bonds, because the risk of loss in the
average single common stock issue is undoubtedly
greater than in the average single bond. But the
comparison has not been true historically of a
diversified group of common stocks, since common
stocks as a whole have had a well-defined upward
bias or long-term upward movement. This in turn is
readily explicable in terms of the country's growth,
plus the steady reinvestment of undistributed prof-
its, plus the strong net inflationary trend since the
turn of the century.
RRE AND CASUALTY RARES
The analogy here is with fire and casualty insur-
ance rates. People pay about twice as much for fire
insurance as their own actuarially determined ex-
posure would indicate--because they cannot
soundly afford to carry the individual risk them-
selves. For similar reasons the overall return on
common stocks appears to have been at least twice
as much as their true overall risk has required. An
interesting relationship at this point appears from
the Keystone chart showing the trend of the Dow-
Jones industrial average since 1899. Both the upper
and lower lines happen to rise at the rate of one
third every ten years. You will recognize this as the
2.90% rate of compound interest realized on U. S.
Savings Bonds, Series E. What this means is the
consistent Dow-Jones investor has obtained the
same increase in principal value as the savings
bonds offer in lieu of interest; and in addition the
Dow-Jones stock investor has obtained all the
annual dividends from his holdings as a bonus
above the Government bond interest rate.
The reasoning I have just indulged in is, I
Financial Analysts Journal / January-February 1995

believe, both scientifically valid and psychologi-
cally dangerous. Its validity depends on the main-
tenance in the stock market of the substantial
disparity between bond yields and the price-earn-
ings ratios on stocks. If as happened in the
1920s--this very thesis is twisted into the slogan
that common stocks are attractive investments,
regardless of h o w high they sell, then we would
find ourselves beginning as scientists and ending
as heedless and ill-starred gamblers. It may be a
fair generalization to assert that the top levels of
most "normal" bull markets are characterized by a
tendency to equate stock risks with bond risks.
These high valuations may indeed have some
justification in pure theory, but the important
thing for us to bear in mind as practicing analysts
is that, when you pay full value for common
stocks, you are in great danger of later appearing
to have paid too much.
INDIVIDUAL UNDERVALUATIONS
Turning now to the field of individual undervalua-
tions, we find ourselves on more familiar ground.
Our work with this group readily admits of the
scientific processes of wide observation and the
testing out of predictions or hypotheses by their
sequels. The theory of undervalued issues must
necessarily require an explanation of their origin.
The explanations are in truth quite varied and
taken together form what may be called a "pathol-
ogy of market prices." They range from obvious
causes, such as an unduly low dividend or a
temporary setback in earnings, to more subtle and
special conditions such as too much common stock
in the capital structure or even too much cash in
the bank. In between lie numerous other causes
such as the presence of important litigation, or the
combination of two dissimilar businesses, or the
use of the now discredited holding company
setup.
ORIGINS OF UNDERVALUATION
UNDERSTOOD
The origins of undervaluation are pretty well un-
derstood by now and could no doubt be set forth
in an acceptably scientific study. We do not know
as much about the cure of undervaluations. In
what proportion of cases is the discrepancy cor-
rected? H o w or w h y does the correction occur?
H o w long does the process take? These questions
remind us somewhat of those we raised about
psychoanalysis at the outset. But one thing of
importance we do know, and that is that the
purchase of undervalued issues on a diversified
Financial Analysts Journal / January-February 1995
1945-1954
basis does produce consistently profitable results.
Thus we have a worthwhile field for more scien-
tific cultivation. Here inductive studies carried on
intelligently and systematically over a period of
years are almost certain to be rewarding.
SELECTION OF GROWTH STOCKS
The third objective of security analysis is the selec-
tion of growth stocks. H o w scientific a procedure
is this now, and how scientific can it be made to
be? Here I enter difficult waters. Most growth
companies are themselves tied in closely with
technological progress; by choosing their shares
the security analyst latches on, as it were, to the
coattails of science. In the 40 or more plant inspec-
tions that are on your scheduled field trips for this
convention week, no doubt your chief emphasis
will be placed on new products and new process
developments; and these in turn will strongly
influence your conclusions about the long-pull
prospects of the various companies. But in most
instances this is primarily a qualitative approach.
Can your work in this field be truly scientific
unless it is solidly based on dependable measure-
ments, that is, specific or minimum projections of
future earnings, and a capitalization of such pro-
jected profits at a rate or multiplier that can be
called reasonably conservative in the light of past
experience? Can a definite price be put on future
growth---below which the stock is a sound pur-
chase, above which it is dear, or in any event
speculative? What is the risk that the expected
growth will fail to materialize? What is the risk of
an important downward change in the market's
evaluation of favorable prospects? A great deal of
systematic study in this field is necessary before
dependable answers to such questions will be
forthcoming.
STOCK INVES'rMENT IN PRESCIENTIFIC
STAGE
In the meantime I cannot help but feel that growth
stock investment is still in the prescientific stage. It
is at the same time more fascinating and less
precise than the selection of safe bonds or under-
valued securities. In the growth stock field, the
concept of margin of safety loses the clarity and the
primacy it enjoys in those other two classes of
security analysis. True, there is safety in growth,
and some of us will go so far as to declare that
there can be no real safety except in growth. But
these sound to me more like slogans than scientif-
ically formulated and verified propositions. Again,
in the growth field the element of selectivity is so
27
1945-1954
prominent as to place diversification in a second-
ary and perhaps dubious position. A case can be
made for putting all your growth eggs in the one
best or a relatively few best baskets. Thus in this
branch of security analysis the actuarial element
may be missing, and that circumstance undoubt-
edly militates against truly scientific procedures
and results.
INVERTED RELATIONSHIP
There is undoubtedly an organic but inverted
relationship between the growth stock con-
cept and the theory of undervalued securities. The
attraction of growth is like a tidal pull which causes
high tides in one area, the assumed growth com-
panies, and low tides in another area, the assumed
nongrowth companies. We can measure, in a
sense, scientifically the distorting effect of this
influence by using as our standard the minimum
business value of enterprises in the nonfavored
group. By way of illustration let us apply that
thought to three California concerns. The shares of
Roos Brothers, a local retail enterprise, will in the
nature of things tend to sell below their analyti-
cally determined value for basically the same rea-
sons that are b o u n d to produce overvaluations in
the shares of Superior Oil or Kern County Land.
I come finally to the standard occupation of
brokerage house analysts and advisory services,
namely, the selection of issues favorably situated
for a near-term market advance. The usual as-
sumption here is that, if the earnings will improve
or the dividend will be raised, then the price will
improve. Thus the process consists essentially of
locating and recommending those companies that
are likely to increase their earnings or dividends in
the near term. You all know the three basic haz-
ards encountered in this work: that the expected
improvement will not take place, that it is already
discounted in the current price, that for some other
reason or for no known reason the price will not
move the w a y it should.
It may be that despite these hazards it is
possible to obtain worthwhile results on the aver-
age from competent short-term analyses and pre-
Foo'rNOTES
1. H.D. Wolfe, "Science as a Trustworthy Tool," The Analysts
Journal (March 1952):45-49.
2. Cogitator, "On Being Right in SecurityAnalysis," The Ana-
28
dictions. Who of us can say whether or not this is
true? In view of the importance of this analytical
work, in terms of time, energy, and money cost, it
might not be a bad idea to subject it to a thorough-
going evaluation.
SEARCHING SELF-EXAMINATION
This brings me to my conclusion and my one
concrete proposal. Security analysis has now
reached the stage where it is ready for a continu-
ous and searching self-examination by the use of
established statistical tools. We should collect the
studies and recommendations of numerous ana-
lysts, classify them in accordance with their objec-
tives (perhaps in the four groups suggested in this
paper), and then do our best to evaluate their
accuracy and success. The purpose of such a
record would not be to show w h o is a good
security analyst and who is a poor one, but rather
to show what methods and approaches are sound
and fruitful and which ones fail to meet the test of
experience.
This suggestion was originally made in the
articles published under the p s e u d o n y m of Cogita-
tor in THE ANALYSTSJOURNALsix years ago. At that
time I wrote: "It is unlikely that security analysis
could develop professional stature in the absence
of reasonably definite and plausible tests of the
soundness of individual and group recommenda-
tions. ''2 The N e w York Society is n o w taking the
first positive steps to establish a quasiprofessional
rating or title for security analysts who meet spec-
ified requirements. It is virtually certain that this
movement will develop ultimately in full-fledged
professional status for our calling. The time may
well be ripe for the Federation and its constituent
Societies to begin a systematic accumulation of
case histories, which should make possible the
transmission of a continuous, ever-growing body
of knowledge and technique from the analysts of
the past to those of the future.
When this work is well under w a y security
analysis may b e g i n - - m o d e s t y , but hopefully--to
refer to itself as a scientific discipline.
lysts Journal (1st Quarter 1946):18-21, especially at p. 18 for
quoted material.
Financial Analysts Journal / January-February 1995